|Publication number||US7467325 B2|
|Application number||US 11/055,258|
|Publication date||Dec 16, 2008|
|Filing date||Feb 10, 2005|
|Priority date||Feb 10, 2005|
|Also published as||US7827443, US20060179207, US20090063898|
|Publication number||055258, 11055258, US 7467325 B2, US 7467325B2, US-B2-7467325, US7467325 B2, US7467325B2|
|Inventors||Susan Elizabeth Eisen, Hung Qui Le, Michael James Mack, Dung Quoc Nguyen, Jose Angel Paredes, Scott Barnett Swaney|
|Original Assignee||International Business Machines Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (76), Non-Patent Citations (7), Referenced by (18), Classifications (17), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present application is related to co-pending application entitled “METHOD FOR CHECKPOINTING INSTRUCTION GROUPS WITH OUT-OF-ORDER FLOATING POINT INSTRUCTIONS IN A MULTI-THREADED PROCESSOR”, Ser. No. 11/054,988, and application entitled “MINI-REFRESH PROCESSOR RECOVERY AS BUG WORKAROUND METHOD USING EXISTING RECOVERY HARDWARE”, Ser. No. 11/055,823.
1. Technical Field
The present invention relates to an improved data processing system. More specifically, the present invention is directed to a method, apparatus, and computer program product for recovering from transient errors in arrays and latches in and supporting a microprocessor by restoring registers to a known correct state earlier checkpointed for the processor and providing for directing processing to a service processor for certain errors.
2. Description of Related Art
A symmetric multiprocessing (SMP) data processing system has multiple processors that are symmetric such that each processor has the same processing speed and latency. An SMP system may be logically partitioned to have one or more operating systems that divide the work into tasks that are distributed evenly among the various processors by dispatching programs to each processor.
Modern micro-processors are usually superscalar, which means a single processor can decode, dispatch, and execute multiple instructions each processor cycle. These modern processors may also support simultaneous multi-threading (SMT), which means each processor can concurrently execute more than one software program (thread) at a time. An SMT processor typically has the ability to favor one thread over another when both threads are running on the same processor. Each thread is assigned a hardware-level priority by the operating system, or by the hypervisor in a logically partitioned environment. The Hypervisor may assist error correction by providing special handling to a microprocessor that has issued a machine check signal or a Hypervisor interrupt.
Static Random Access Memories (SRAM) have been susceptible to transient errors due to naturally occurring radiation for several generations of integrated circuits. As the scale of gates of various kinds has been reduced, even non-SRAMs, e.g. latches, have become susceptible to this problem. This phenomenon must be handled in order for further reduced size architectures to be useful and always correct when delivered to a customer in a processing device.
Also potentially problematic is contending with extremely rare sequences and combinations of instructions and states that invariably result in incorrect results each time such sequences and combinations occur. Typically, such so called ‘functional errors’ or ‘bugs’ would be discovered through intensive testing of a design prior to general availability. With extremely complex, superscalar, multi-threaded processors, used in incrementally scaleable large SMPs, with large numbers of virtual partitions, the verification state space approaches infinite. Validation of such a large state space often exceeds the capacity of formal verification tools and simulation test cases. Prototype hardware is typically manufactured for intensive testing at machine speeds, but unfortunately some mis-handled combinations of rare events may occur so infrequently that they are encountered very late or not at all during the prototype testing. Modifying and manufacturing additional prototypes to fix late found design bugs is expensive and time consuming, which may delay a product from reaching the market.
Often such design errors could be avoided by reducing the number and complexity of operations going on in the processor, thereby dramatically reducing the total state space, making the mis-handled combination of events more rare, or even impossible. Avoiding the use of complex superscalar pipelining techniques such as multiple instruction decode, dispatch, and execution; load and branch look-aheads; imprecise exception mode; pre-fetching; out-of-order processing, and simultaneous multi-threading (SMT) would reduce the total possible state space of a processor to a level where simulation tools would be adequate to ensure correct operation. However, modern processor throughput demands are such that dropping such techniques entirely would result in a commercially unviable processor. But it would be advantageous to temporarily suspend or disable such complex controls only when required to avoid a mis-handled combination of rare events. It is unknown in the prior art to forbear from using superscalar pipelining techniques and other modes (now considered normal) just so that a sequence of instructions which encounters erroneous operation can be retried successfully by avoiding combinations of rare events which resulted in the erroneous operation.
Increasing circuit density with new technologies is causing power consumption to become a limiting factor in microprocessor designs. In order to minimize power consumption, portions of the circuitry which are not required for a particular active operation are “turned off” by suppressing the clocks to them. Suppressing the clocks results in less circuit switching, and hence less power consumption. During periods of very low workload, large portions of the processor may be put into a low-power state, sometimes referred to as “nap” or “doze” modes. In the event of an error, where a prior checkpoint state is refreshed to the processor, the logic which is in the low-power state must be woken to allow it to also be reset and refreshed to the prior checkpoint state. This management of low-power states during processor recovery is not included in the prior art.
Virtualization of processors in large SMP systems requires efficient (fast) address translation to maintain throughput. A common technique to improve address translation performance is through the use of “look-aside” buffers which remember results from prior translations so they can be simply reused instead of recalculated. A look-aside buffer contains a relatively small number of entries, so after some time entries need to be discarded to make room for newer entries. If the result for a translation is not available in a look-aside buffer, it must be re-calculated through a series of memory accesses and additions. Once the first pointer to memory is known, hardware state machines can traverse a linked-list of address pointers to perform the translation. However, the first address pointer, which points to a storage “segment”, cannot be determined by the hardware state machines. Segment pointers are managed by the operating system and hypervisor, and are stored in a Segment Lookaside Buffer (SLB) in the processor.
Unfortunately, the size of the SLB is such that it is prohibitively expensive to provide a backed-up copy of it within the processor chip die. Thus, in the event of any failure, a means to determine if the SLB contents were potentially corrupted and obtain and synchronize backed-up data is necessary, but not yet encountered in the prior art.
According to a preferred embodiment, the present invention operates on a processor core in a chip of a symmetric multiprocessing system, the chip having at least one processing core and a host firmware. A fault occurs during the processor core execution and is detected by error detection circuitry. Recovery circuits are woken from a low-power state by, e.g. clocking them. In addition, all other low-power circuits of the processor core are woken. Checkpointing of processor core state ceases as a determination is made if the state meets criteria for continued processing by the invention. The processor core is logically removed from the system, and reset. Values from the most recent checkpoint are stored back to all registers and architected facilities required to resume processing from the checkpoint. Logic which is in a low-power (e.g. clock off) state is woken (e.g. clocked) so that it can be reset and restored to the prior checkpoint state. The processor core is restored to the system. A high priority interrupt may occur in some instances, as the processor core is allowed to resume i-fetching and dispatch, albeit in a reduced execution mode. After a limited number of instruction groups process without error, the processor core is allowed to resume operating in normal mode.
The novel features believed characteristic of the invention are set forth in the appended claims. The invention itself, however, as well as a preferred mode of use, further objectives and advantages thereof, will best be understood by reference to the following detailed description of an illustrative embodiment when read in conjunction with the accompanying drawings, wherein:
Embodiments of the invention may provide an opportunity for a Hypervisor or firmware to remedy an error in a processor core when efforts of circuits in the chip housing the processor core have been exhausted.
We also provide an advantage in that we allow embodiments to retry instructions in a safer and more robust mode to avoid a functional error known to occur when operating in normal mode.
The present invention is a method, apparatus, and computer program product for handling errors in a processor core by using backed up register and buffer values in place of data known to be, at least in part, in error, e.g. as caused by random incident radiation.
With reference now to the figures,
Also connected to System Bus 106 is Memory Controller/Cache 108, which provides an interface to Local Memory 109. Input/Output (I/O) Bus Bridge 110 is connected to System Bus 106 and provides an interface to I/O Bus 112. Memory Controller/Cache 108 and I/O Bus Bridge 110 may be integrated as depicted.
Peripheral Component Interconnect (PCI) Bus Bridge 114 connected to I/O Bus 112 provides an interface to PCI Local Bus 116. A number of modems may be connected to PCI Bus 116. Typical PCI Bus implementations will support four PCI expansion slots or add-in connectors. Communications links to network computers may be provided through Modem 118 and Network Adapter 120 connected to PCI Local Bus 116 through add-in boards.
Network Adapter 120 includes a Physical Layer 182 which conditions analog signals to go out to the network, such as for example an Ethernet network over an R45 connector. A Media Access Controller (MAC) 180 is included within Network Adapter 120. MAC 180 is coupled to Bus 116 and processes digital network signals. MAC 180 serves as an interface between Bus 116 and Physical Layer 182. MAC 180 performs a number of functions involved in the transmission and reception of data packets. For example, during the transmission of data, MAC 180 assembles the data to be transmitted into a packet with address and error detection fields. Conversely, during the reception of a packet, MAC 180 disassembles the packet and performs address checking and error detection. In addition, MAC 180 typically performs encoding/decoding of digital signals transmitted and performs preamble generation/removal as well as bit transmission/reception.
Additional PCI Bus Bridges 122 and 124 provide interfaces for additional PCI Buses 126 and 128, from which additional modems or network adapters may be supported. In this manner, data processing system 100 allows connections to multiple network computers. A memory-mapped Graphics Adapter 130 and Hard Disk 132 may also be connected to I/O Bus 112 as depicted, either directly or indirectly.
Service Processor 104 interrogates system processors, memory components, and I/O bridges to generate an inventory and topology understanding of data processing system 100. Service Processor 104 also executes Built-In-Self-Tests (BISTs), Basic Assurance Tests (BATs), and memory tests on all elements found by interrogating a system processor, memory controller, and I/O bridge. Any error information for failures detected during the BISTs, BATs, and memory tests are gathered and reported by Service Processor 104.
System Bus 106 can be connected to one or more like system busses which allows the data processing system to be incrementally scaled up to a large n-way SMP.
Those of ordinary skill in the art will appreciate that the hardware depicted in
As shown in
In a preferred embodiment of the present invention, the execution circuitry of the processor may include, Branch Unit 206, Fixed-Point Execution Units 208 a (FXUA) and 208 b (FXUB), Load/Store Units 207 a (LSUA) and 207 b (LSUB), and Floating-Point Execution Units 209 a (FPUA) and 209 b (FPUB). Execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b are fully shared across both threads. The processor includes multiple register sets 210 a, 210 b, 211 a, 211 b, 212 a, 212 b, and 212 c separately for each of the two threads, namely General Purpose Registers 210 a and 210 b (GPR), Floating-Point Registers 211 a and 211 b (FPR), and Special Purpose Registers 212 a and 212 b (SPR). The processor additionally includes a set of SPRs 212 c which is shared across both threads. Simplified internal bus structure 217 is shown to depict connections between execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b and register sets 210 a, 210 b, 211 a, 211 b, 212 a, 212 b, and 212 c.
FPUA 209 a and FPUB 209 b input their register source operand information from and output their destination register operand data to FPRs 211 a and 211 b according to which thread each executing instruction belongs to. FXUA 208 a, FXUB 208 b, LSUA 207 a, and LSUB 207 b input their register source operand information from and output their destination register operand data to GPRs 210 a and 210 b according to which thread each executing instruction belongs to. A subset of instructions executed by FXUA 208 a, FXUB 208 b, and Branch Unit 206 use SPRs 212 a, 212 b and 212 c as source and destination operand registers. LSUA 207 a and LSUB 207 b input their storage operands from and output their storage operands to Data Cache 202 which stores operand data for multiple programs (threads). Data Cache 202 also has an interface to Level 2 Cache/Memory 220.
Level 2 cache 202 may also have (not shown) associated with it a non-cacheable unit (NCU) which accepts load/stores from the processor and forwards directly to memory 220, thus bypassing the coherency protocols required for storage to cache. Memory 220 could also be memory-mapped registers or resource.
In response to the instructions input from Instruction Cache 201 and decoded by Instruction Decode Unit 204, IDU 205 selectively dispatches the instructions to execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b. Execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b execute one or more instructions of a particular class of instructions. For example, FXUA 208 a and FXUB 208 b execute fixed-point mathematical operations on register source operands, such as addition, subtraction, ANDing, ORing and XORing. FPUA 209 a and FPUB 209 b execute floating-point mathematical operations on register source operands, such as floating-point multiplication and division. LSUA 207 a and LSUB 207 b execute load and store instructions which move operand data between Data Cache 202 and registers 210 a, 210 b, 211 a, and 211 b. Branch Unit 206 executes branch instructions which conditionally alter the flow of execution through a program by modifying the instruction address used by IFU 203 to request instructions from Instruction Cache 201.
IDU 205 groups together decoded instructions to be executed at the same time, depending on the mix of decoded instructions and available execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b to perform the required operation for each instruction. For example, because there are only two Load/Store Units 207 a and 207 b, a maximum of two Load/Store type instructions may be grouped together. In a preferred embodiment of the present invention, up to seven instructions may be grouped together (two Fixed-Point arithmetic, two Load/Store, two Floating-Point arithmetic, and one Branch), and up to five instructions may belong to the same thread. IDU 205 includes in the group as many instructions as possible from the higher priority thread, up to five, before including instructions from the lower priority thread. Values in Special Purpose Registers 212 a and 212 b indicate thread priority 218 to IDU 205.
Instruction Completion Unit (ICU) 216 monitors internal bus structure 217 to determine when instructions executing in execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b are finished writing their operand results. Instructions executed by Branch Unit 206, FXUA 208 a, FXUB 208 b, LSUA 207 a, and LSUB 207 b require the same number of cycles to execute, while instructions executed by FPUA 209 a and FPUB 209 b require a variable, and a larger number of cycles to execute. Therefore, instructions which are grouped together and start executing at the same time do not necessarily finish executing at the same time. When all the instructions grouped together within a thread are finished writing their operand results, the group is said to be “completed”.
Instruction Completion Unit 216 monitors for the completion of instruction, and sends control information 219 to IDU 205 to identify that more groups of instructions can be dispatched to execution units 206, 207 a, 207 b, 208 a, 208 b, 209 a, and 209 b. IDU 205 sends control information 225 to IFU 203 and Instruction Decode Unit 204 to indicate that it is ready to receive more decoded instructions.
Processor 200 preferable also includes error detection circuitry (not shown on
The register results are held in Write Queue 213 a and 213 b and FPR Reorder Buffer 214 a and 214 b for each thread until Completion Unit 216 indicates (via control 219) that all instructions in the group have completed for each thread. If no errors are reported by the error detection circuitry (not shown in
Some of SPRs 212 a, 212 b, and 212 c hold information used to control the operation of the various functional units. This is represented, simplified, in
Recovery unit 215 includes control logic (not shown in
Scattered among the units and registers of processor core 200, are a number of circuits which implement functions known as pervasives, which includes (but not limited to) controlling clocking for low-power states, collection of reported errors, interface to service processor, and control signals to override the normal superscalar, pipelined, Simultaneous Multi-Threading (SMT) operation of their host units and placing their respective hosts into a recovery mode.
Because store conditional (STCX) and cache-inhibited load (CI-Load) are long duration instructions which modify state outside of the processor core prior to their completion, once started they must be allowed to complete and reach the checkpointed state. Thus it may be necessary to wait several clock cycles for off-chip circuits to respond with signals concerning the completion of such instructions. Thus an affirmative branch of the active long duration instruction (step 311) by the recovery unit causes the recovery unit to allow the advancement of the checkpoint (step 313) to include the STCX or CI-Load instruction. Instruction grouping rules in the instruction dispatch unit (205 from
Since, for long duration instructions, STCX and CI-Load, the checkpoint may be allowed to advance after an error is detected, errors which may affect the results of such instructions are isolated from other errors which would not. In the context of error recovery, the term ‘may’ means there is a possibility, generally uncontrollable by the user or device, that the error produce an incorrect result. The uncertainty can be caused by faulty error detection, rather than a fault in the tested circuit itself, among other things. Recovery unit 215 of
Similarly, had there been an absence of STCX or CI-load operations active (step 311), execution of the recovery routine would flow directly to checkpoint blocking (step 317) of the processor core.
Recovery unit 215 may wait for the write queue to be drained up to the checkpoint boundary. Once recovery unit 215 advances the checkpoint and has a group tag (gtag) which identifies the instruction at the boundary of the checkpoint, recovery unit sends the gtag to ICU 216. ICU has access to a next instruction address (NIA) and may discover an error—thus signaling the status to recovery unit. Recovery unit may checkstop 399 since a faulty NIA may be an escalation condition (step 331). Other conditions may also trigger an escalation (step 331) which aborts recovery ending in checkstop 399. Errors which may compromise the integrity of the checkpointed values or the NIA are isolated from other errors so that they can be observed by decision block (step 331).
In very generalized terms, a “quiescing” is an attempt to slow and stop various side processes so that, in this case, recovery can progress with the least amount of confusion, i.e. to temporarily disable, even by letting processes expire. In the absence of an escalation condition, recovery unit may wait a timeout period to detect a quiesce occurrence of the NCU and L2. The recovery sequence suspends operation of the processor for some time, so to ensure no operations elsewhere in the SMP system are dependent on a response from the processor, the interface to the L2 and NCU is quiesced.
A quiesce condition (step 351), also known as quiescing the cache functional unit, may occur following five events: 1) LSU 207 allows any checkpointed store data remaining in its store queue to drain past the private L1 data cache 202 to the shared L2 cache; 2) L2 writes all store data sent from the processor into the cache; 3) any outstanding memory access (fetch) requests from the processor finish, but are dropped; 4) processor signals the L2 and NCU to stop sending any new invalidate commands which require responses from the processor; and 5) processor sends an invalidate response in case there was an active invalidate command waiting for a response. After these 5 steps, the L2 and NCU should not be waiting for any response from the processor or processing any operations on behalf of the processor, so is said to be idle. The idle condition is indicated to the processor. Gaps between commands during periods of low activity can result in occasional cycles where the interface is idle, but once idle for many (e.g. 256) consecutive cycles, the interface is said to be quiesced. Failure to detect a quiesced condition within the preprogrammed timeout results in checkstop 399. Otherwise recovery attempts to continue unless a recovery count threshold is exceeded (step 341), which may result in a checkstop 399. The recovery count threshold is not checked until after the coherent checkpoint has been established and the L2 and NCU interfaces quiesced, as waiting for this state allows for easier system-level recovery actions, such as alternate processor recovery, which are beyond the scope of this invention.
Absent a recovery count threshold being exceeded, pervasives circuits may logically remove the processor core from the system (step 361). This step may occur by gating off (fencing) the interface to L2 cache, NCU and memory 220 of
Extensive prototype testing may identify specific error conditions which require Service Processor firmware to aid in the recovery process by overriding settings or parameters not directly controlled by the recovery state machine circuitry. If any such cases are found to exist, a configuration setting may be selected at boot time to indicate service processor intervention is required. At this time, if service processor intervention is required (step 371) an interrupt is presented to the Service Processor by the recovery logic. The service processor (SP) may then observe the state of various registers and latches throughout the processor via scan communication (SCOM), or level sensitive scan design (LSSD) scanning, to determine which if any predetermined circumstances exist which require additional support. If required, SP takes over processing (step 372) by means known in the art. Scan communication techniques are known from US patent application publication No. US 2004/0215929 “Cross-chip communication mechanism in distributed node topology”.
If extensive testing shows that service processor intervention is not required for any special cases, the interrupt may be disabled such that decision (step 371) is never answered yes. Once the service processor completes its handling of the condition, or if none was required, the recovery unit may reset the contents of arrays in the processor, such as the instruction fetch cache 201 and the data cache 202, by using Array Built-In Self Test (ABIST) circuitry integrated with each array design. All arrays in the processor except for the segment look-aside buffer are reset. Collectively the resetting of arrays is known as array reset (step 383), and must complete within a preprogrammed time period or the recovery unit will place the core processor in checkstop 399. Alternatively, the recovery logic may reset latches in the processor core. Because there is no backup for the SLB which is directly accessible by the recovery logic, the SLB is left intact. Errors which may corrupt the SLB contents must be isolated from other errors, and the corruption indicated to firmware and operating system when processing resumes. An error is a state where one or more latches, registers or other storage devices contain an incorrect datum, or where a switching device, e.g. a transistor, fails to switch within a specified time of an input reaching a specified value, or a circuit opening or closing due to mechanical, or other forces well known in the art as broken.
Roughly concurrent with the array reset, logic resets are activated to restore various logic circuits to a known default state (step 381). Some logic circuits may reset in stages or concurrently. Various mechanisms for resetting logic latches are known, not limited to asynchronous latch reset inputs, activating LSSD scan clocks to “flush” a constant value through scan chains, combinatorially gating a value on data inputs, and initiating a state machine to provide pre-determined values or select values from an alternate source. Our preferred embodiment uses a combination of such known mechanism for resetting logic latches to a known error-free state. Error collection registers which combine reported errors from throughout the processor to report them to the recovery logic are also cleared by the logic reset.
Another opportunity for the SP to conduct special handling occurs (step 385). Again, if intervention may be required, an interrupt is presented to the SP. The SP analyzes the processor state via SCOM or scan operations, and applies any necessary predetermined modifications (step 386).
Errors detected during the writing of each register from its checkpoint copy cause recovery to abort to a checkstop. Successful completion of writing checkpoint values to SPRs, FPRs, and GPRs will result in restoring (step 391) the processor core to the system, e.g. by removing the gating (fencing) from the interface to L2 cache, NCU and memory 220 of
IFU may restart instruction fetching (step 392) using the next instruction address (NIA) register. At this time, all formerly low-power circuits are allowed to return to the low-power state by discontinuing the overriding of clocks (step 393). Although it is likely that some number of these circuits will again be woken by a HMI or machine check interrupt being taken.
Avoidance of functional errors may occur at this point. A number of performance lowering execution modes (step 394) may be selected, which reduces the degree of complex superscalar multi-threaded pipelined instruction processing. Examples include, but are not limited to: single instruction dispatch; disable load look-aheads; disable branch look-aheads; precise exception mode (in-order serialized floating point execution) and disable pre-fetches. Collectively, the one or several modes that are selected comprise a reduced execution mode. At least one configuration register may select which modes are activated, and these predetermined modes may be set during machine initialization. A preprogrammed number of new groups of instructions are executed and checkpointed in the reduced execution mode. The number serves as a threshold. It should be noted that reduced execution mode has meaning only in relation to the normal execution mode, in the sense that the reduced execution mode adopts a processing mode that is generally slower alternative to one used in the normal execution mode, in relation to significant tracts of instructions executed over time. It is noted that there are occasional instances where branch-lookahead actually executes fewer instructions than a processor executing the same instructions without branch-lookahead as occurs when the prediction of a branch determination is guessed wrong by the processor. However, under long stretches of time, using an ordinary and routine set of instructions, a processor with branch-lookahead statistically outperforms the processor without branch-lookahead, all other computing resources being equal. Nevertheless, branch lookahead is considered a faster alternative to executing without branch lookahead. The same holds true for each of multiple instruction dispatch; load look-aheads; imprecise exception mode and pre-fetching-as compared to their mutually exclusive alternative modes.
Errors detected (step 395) after this point are handled by re-executing the steps that follow from the link “A” 302 on
The final restorative steps may include clearing (step 388) the recovery count of the “recovery count threshold exceeded” (step 341) and “checkpoint count reached” (step 397) respectively. Also, normal execution mode (step 398) resumes, wherein normal execution mode may include one or more of the following: multiple instruction dispatch; load look-aheads; branch look-aheads; imprecise exception mode and pre-fetching or any other mode that purports, in the long run, to operate faster than a mutually exclusive alternative mode. At least one instruction group is executed in the normal execution mode.
The checkpoint count and checkpoint threshold, although not depicted in such detail in the flow diagram
Thus we show how attempts may be made by a processor core and supporting circuitry to, among other things, drop into a reduced execution mode for a certain number of instructions and see if such attempts are sufficiently robust to avoid a rare line-up of events such that an error previously accounted with a group of instructions is avoided.
It is appreciated that many other modes may exist or be discovered that enable a processor to operate in a more aggressive, and potentially error prone mode. Collectively all such equivalents are normal execution modes.
It is important to note that while the present invention has been described in the context of a fully functioning data processing system, those of ordinary skill in the art will appreciate that the processes of the present invention are capable of being distributed in the form of a computer readable medium of instructions and a variety of forms and that the present invention applies equally regardless of the particular type of signal bearing media actually used to carry out the distribution. Examples of computer readable media include recordable-type media, such as a floppy disk, a hard disk drive, a RAM, CD-ROMs, DVD-ROMs, and transmission-type media, such as digital and analog communications links, wired or wireless communications links using transmission forms, such as, for example, radio frequency and light wave transmissions. The computer readable media may take the form of coded formats that are decoded for actual use in a particular data processing system.
The description of the present invention has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiment was chosen and described in order to best explain the principles of the invention, the practical application, and to enable others of ordinary skill in the art to understand the invention for various embodiments with various modifications as are suited to the particular use contemplated.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3736566||Aug 18, 1971||May 29, 1973||Ibm||Central processing unit with hardware controlled checkpoint and retry facilities|
|US4493035||Dec 7, 1982||Jan 8, 1985||Motorola, Inc.||Data processor version validation|
|US4594710 *||Dec 15, 1983||Jun 10, 1986||Fujitsu Limited||Data processing system for preventing machine stoppage due to an error in a copy register|
|US4912707||Aug 23, 1988||Mar 27, 1990||International Business Machines Corporation||Checkpoint retry mechanism|
|US5040107 *||Jul 3, 1989||Aug 13, 1991||International Computers Limited||Pipelined processor with look-ahead mode of operation|
|US5241636 *||Feb 14, 1990||Aug 31, 1993||Intel Corporation||Method for parallel instruction execution in a computer|
|US5345583 *||Apr 5, 1993||Sep 6, 1994||Scientific-Atlanta, Inc.||Method and apparatus for momentarily interrupting power to a microprocessor to clear a fault state|
|US5361267 *||Apr 24, 1992||Nov 1, 1994||Digital Equipment Corporation||Scheme for error handling in a computer system|
|US5418916||Oct 4, 1990||May 23, 1995||International Business Machines||Central processing unit checkpoint retry for store-in and store-through cache systems|
|US5423026 *||Sep 5, 1991||Jun 6, 1995||International Business Machines Corporation||Method and apparatus for performing control unit level recovery operations|
|US5446851||Jun 8, 1994||Aug 29, 1995||Matsushita Electric Industrial Co., Ltd.||Instruction supplier for a microprocessor capable of preventing a functional error operation|
|US5452437||Nov 18, 1991||Sep 19, 1995||Motorola, Inc.||Methods of debugging multiprocessor system|
|US5467473||Jan 8, 1993||Nov 14, 1995||International Business Machines Corporation||Out of order instruction load and store comparison|
|US5495587||Jun 21, 1994||Feb 27, 1996||International Business Machines Corporation||Method for processing checkpoint instructions to allow concurrent execution of overlapping instructions|
|US5502826||Jan 25, 1994||Mar 26, 1996||International Business Machines Corporation||System and method for obtaining parallel existing instructions in a particular data processing configuration by compounding instructions|
|US5551043||Sep 7, 1994||Aug 27, 1996||International Business Machines Corporation||Standby checkpoint to prevent data loss|
|US5590277||Jun 22, 1994||Dec 31, 1996||Lucent Technologies Inc.||Progressive retry method and apparatus for software failure recovery in multi-process message-passing applications|
|US5630075||May 25, 1995||May 13, 1997||Intel Corporation||Write combining buffer for sequentially addressed partial line operations originating from a single instruction|
|US5664137 *||Sep 7, 1995||Sep 2, 1997||Intel Corporation||Method and apparatus for executing and dispatching store operations in a computer system|
|US5692121||Apr 30, 1996||Nov 25, 1997||International Business Machines Corporation||Recovery unit for mirrored processors|
|US5737604 *||Sep 30, 1996||Apr 7, 1998||Compaq Computer Corporation||Method and apparatus for independently resetting processors and cache controllers in multiple processor systems|
|US5748873 *||Jul 26, 1996||May 5, 1998||Hitachi,Ltd.||Fault recovering system provided in highly reliable computer system having duplicated processors|
|US5751985||Jun 7, 1995||May 12, 1998||Hal Computer Systems, Inc.||Processor structure and method for tracking instruction status to maintain precise state|
|US5812757 *||Jul 5, 1996||Sep 22, 1998||Mitsubishi Denki Kabushiki Kaisha||Processing board, a computer, and a fault recovery method for the computer|
|US5826070||Aug 30, 1996||Oct 20, 1998||International Business Machines Corporation||Apparatus and method for maintaining status flags and condition codes using a renaming technique in an out of order floating point execution unit|
|US5867444 *||Sep 25, 1997||Feb 2, 1999||Compaq Computer Corporation||Programmable memory device that supports multiple operational modes|
|US5872948 *||Mar 15, 1996||Feb 16, 1999||International Business Machines Corporation||Processor and method for out-of-order execution of instructions based upon an instruction parameter|
|US5875326||Apr 25, 1997||Feb 23, 1999||International Business Machines Corporation||Data processing system and method for completing out-of-order instructions|
|US5884061||May 1, 1995||Mar 16, 1999||International Business Machines Corporation||Apparatus to perform source operand dependency analysis perform register renaming and provide rapid pipeline recovery for a microprocessor capable of issuing and executing multiple instructions out-of-order in a single processor cycle|
|US5892978||Jul 24, 1996||Apr 6, 1999||Vlsi Technology, Inc.||Combined consective byte update buffer|
|US5923832||Mar 14, 1997||Jul 13, 1999||Kabushiki Kaisha Toshiba||Method and apparatus for checkpointing in computer system|
|US5978896||Aug 12, 1994||Nov 2, 1999||International Business Machines Corporation||Method and system for increased instruction dispatch efficiency in a superscalar processor system|
|US5996083||Aug 11, 1995||Nov 30, 1999||Hewlett-Packard Company||Microprocessor having software controllable power consumption|
|US6058491||Sep 15, 1997||May 2, 2000||International Business Machines Corporation||Method and system for fault-handling to improve reliability of a data-processing system|
|US6289428 *||Aug 3, 1999||Sep 11, 2001||International Business Machines Corporation||Superscaler processor and method for efficiently recovering from misaligned data addresses|
|US6360333 *||Nov 19, 1998||Mar 19, 2002||Compaq Computer Corporation||Method and apparatus for determining a processor failure in a multiprocessor computer|
|US6363475||Aug 31, 2000||Mar 26, 2002||Micron Technology, Inc.||Apparatus and method for program level parallelism in a VLIW processor|
|US6393582 *||Dec 10, 1998||May 21, 2002||Compaq Computer Corporation||Error self-checking and recovery using lock-step processor pair architecture|
|US6415376||Jun 16, 2000||Jul 2, 2002||Conexant Sytems, Inc.||Apparatus and method for issue grouping of instructions in a VLIW processor|
|US6438709 *||Dec 29, 1997||Aug 20, 2002||Intel Corporation||Method for recovering from computer system lockup condition|
|US6543002 *||Nov 4, 1999||Apr 1, 2003||International Business Machines Corporation||Recovery from hang condition in a microprocessor|
|US6553480||Nov 5, 1999||Apr 22, 2003||International Business Machines Corporation||System and method for managing the execution of instruction groups having multiple executable instructions|
|US6571324 *||Jun 26, 1997||May 27, 2003||Hewlett-Packard Development Company, L.P.||Warmswap of failed memory modules and data reconstruction in a mirrored writeback cache system|
|US6625749 *||Dec 21, 1999||Sep 23, 2003||Intel Corporation||Firmware mechanism for correcting soft errors|
|US6640313 *||Dec 21, 1999||Oct 28, 2003||Intel Corporation||Microprocessor with high-reliability operating mode|
|US6718483 *||Jul 21, 2000||Apr 6, 2004||Nec Corporation||Fault tolerant circuit and autonomous recovering method|
|US6738926||Jun 15, 2001||May 18, 2004||Sun Microsystems, Inc.||Method and apparatus for recovering a multi-threaded process from a checkpoint|
|US6751756 *||Dec 1, 2000||Jun 15, 2004||Unisys Corporation||First level cache parity error inject|
|US6834358 *||Mar 28, 2001||Dec 21, 2004||Ncr Corporation||Restartable database loads using parallel data streams|
|US6948092 *||Jan 31, 2002||Sep 20, 2005||Hewlett-Packard Development Company, L.P.||System recovery from errors for processor and associated components|
|US7055060 *||Dec 19, 2002||May 30, 2006||Intel Corporation||On-die mechanism for high-reliability processor|
|US7096322 *||Oct 10, 2003||Aug 22, 2006||Unisys Corporation||Instruction processor write buffer emulation using embedded emulation control instructions|
|US7124224 *||Dec 22, 2000||Oct 17, 2006||Intel Corporation||Method and apparatus for shared resource management in a multiprocessing system|
|US7158520||Mar 22, 2002||Jan 2, 2007||Juniper Networks, Inc.||Mailbox registers for synchronizing header processing execution|
|US7180893||Mar 22, 2002||Feb 20, 2007||Juniper Networks, Inc.||Parallel layer 2 and layer 3 processing components in a network router|
|US7212530||Mar 22, 2002||May 1, 2007||Juniper Networks, Inc.||Optimized buffer loading for packet header processing|
|US7215662||Mar 22, 2002||May 8, 2007||Juniper Networks, Inc.||Logical separation and accessing of descriptor memories|
|US7236501||Mar 22, 2002||Jun 26, 2007||Juniper Networks, Inc.||Systems and methods for handling packet fragmentation|
|US7239630||Mar 22, 2002||Jul 3, 2007||Juniper Networks, Inc.||Dedicated processing resources for packet header generation|
|US7283528||Mar 22, 2002||Oct 16, 2007||Raymond Marcelino Manese Lim||On the fly header checksum processing using dedicated logic|
|US7392267||Dec 4, 2002||Jun 24, 2008||International Business Machines Corporation||Annotation validity using partial checksums|
|US20010042198 *||Dec 29, 1997||Nov 15, 2001||David I. Poisner||Method for recovering from computer system lockup condition|
|US20030014736 *||Jul 15, 2002||Jan 16, 2003||Nguyen Tai H.||Debugger breakpoint management in a multicore DSP device having shared program memory|
|US20030061535 *||Sep 21, 2001||Mar 27, 2003||Bickel Robert E.||Fault tolerant processing architecture|
|US20030163763||Feb 27, 2002||Aug 28, 2003||Eric Delano||Checkpointing of register file|
|US20040024994||Jul 31, 2002||Feb 5, 2004||International Business Machines Corporation||Method and system for identifying instruction completion delays in a processor|
|US20050044311||Aug 22, 2003||Feb 24, 2005||Oracle International Corporation||Reducing disk IO by full-cache write-merging|
|US20050050304||Aug 29, 2003||Mar 3, 2005||Mukherjee Shubhendu S.||Incremental checkpointing in a multi-threaded architecture|
|US20050149769 *||Dec 29, 2003||Jul 7, 2005||Intel Corporation||Methods and apparatus to selectively power functional units|
|US20060020851 *||Nov 10, 2004||Jan 26, 2006||Fujitsu Limited||Information processing apparatus and error detecting method|
|US20060047958 *||Aug 25, 2004||Mar 2, 2006||Microsoft Corporation||System and method for secure execution of program code|
|US20060143509 *||Dec 20, 2004||Jun 29, 2006||Sony Computer Entertainment Inc.||Methods and apparatus for disabling error countermeasures in a processing system|
|US20060156177||Dec 29, 2004||Jul 13, 2006||Sailesh Kottapalli||Method and apparatus for recovering from soft errors in register files|
|US20070157138||Feb 6, 2007||Jul 5, 2007||The Mathworks, Inc.||Management of functions for block diagrams|
|US20070183425||Apr 3, 2007||Aug 9, 2007||Juniper Networks, Inc.||Logical separation and accessing of descriptor memories|
|USH1291 *||Dec 20, 1990||Feb 1, 1994||Hinton Glenn J||Microprocessor in which multiple instructions are executed in one clock cycle by providing separate machine bus access to a register file for different types of instructions|
|1||*||Banātre et al., "An Architecture for Tolerating Processor Failures in Shared-Memory Multiprocessors", IEEE Transactions on Computers, vol. 45, No. 10, pp. 1101-1115, Oct. 1996.|
|2||Bishop et al., Method for Checkpointing Instruction Groups with Out-of-Order Floating Point Instructions in a Multi-Threaded Processor, Feb. 10, 2005.|
|3||Chained Store Operations for Improved Main Store Performance, IBM Technical Disclosure Bulletin, Sep. 1990, US, vol. 33, issue 4, p. 284-286, NN9009284.|
|4||Floyd et al., Mini-Refresh Processor Recovery as Bug Workaround Method Using Existing Recovery Hardware, Feb. 10, 2005.|
|5||*||Janssens, Bob and Fuchs, W. K., "The Performance of Cache-Based Error Recovery in Multiprocessors", IEEE Transactions on Parallel and Distributed Systems, vol. 5, No. 10, pp. 1033-1043, Oct. 1994.|
|6||Multiprocessing Checkpoint Retry, IBM Technical Disclosure Bulletin, Jul. 1971, vol. 14, issue 2, p. 358-361, NN7107358.|
|7||*||Siewiorek, Daniel P., "Architecture of Fault-Tolerant Computers: An Historical Perspective", The Proceedings of the IEEE, vol. 79, No. 12, pp. 1710-1734, Dec. 1991.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7966536 *||Apr 11, 2008||Jun 21, 2011||International Business Machines Corporation||Method and apparatus for automatic scan completion in the event of a system checkstop|
|US8020039 *||Dec 1, 2008||Sep 13, 2011||Arm Limited||Recovering from errors in streaming DSP applications|
|US8335881||Mar 26, 2010||Dec 18, 2012||Freescale Semiconductor, Inc.||Method and apparatus for handling an interrupt during testing of a data processing system|
|US8438442 *||Mar 26, 2010||May 7, 2013||Freescale Semiconductor, Inc.||Method and apparatus for testing a data processing system|
|US8977895 *||Jul 18, 2012||Mar 10, 2015||International Business Machines Corporation||Multi-core diagnostics and repair using firmware and spare cores|
|US8984335 *||Mar 15, 2013||Mar 17, 2015||International Business Machines Corporation||Core diagnostics and repair|
|US9316737||May 21, 2013||Apr 19, 2016||Spireon, Inc.||Container verification through an electrical receptacle and plug associated with a container and a transport vehicle of an intermodal freight transport system|
|US9342395||Sep 30, 2014||May 17, 2016||International Business Machines Corporation||Error checking using serial collection of error data|
|US9348686||Mar 14, 2014||May 24, 2016||International Business Machines Corporation||Error checking using serial collection of error data|
|US9436786 *||Aug 4, 2015||Sep 6, 2016||Xilinx, Inc.||Method and circuits for superclocking|
|US20090150722 *||Dec 1, 2008||Jun 11, 2009||Alastair David Reid||Recovering from errors in streaming dsp applications|
|US20090259899 *||Apr 11, 2008||Oct 15, 2009||Ralf Ludewig||Method and apparatus for automatic scan completion in the event of a system checkstop|
|US20100205477 *||Mar 17, 2010||Aug 12, 2010||Sony Computer Entertaintment Inc.||Memory Handling Techniques To Facilitate Debugging|
|US20110238878 *||Mar 26, 2010||Sep 29, 2011||Welguisz David M||Method and apparatus for handling an interrupt during testing of a data processing system|
|US20110239070 *||Mar 26, 2010||Sep 29, 2011||Morrison Gary R||Method and apparatus for testing a data processing system|
|US20120233499 *||Mar 6, 2012||Sep 13, 2012||Thales||Device for Improving the Fault Tolerance of a Processor|
|US20140025991 *||Jul 18, 2012||Jan 23, 2014||International Business Machines Corporation||Core diagnostics and repair|
|US20140108859 *||Mar 15, 2013||Apr 17, 2014||International Business Machines Corporation||Core diagnostics and repair|
|U.S. Classification||714/10, 711/E12.021, 712/E09.053, 712/227, 712/E09.061, 712/E09.033|
|Cooperative Classification||G06F9/3851, G06F11/1407, G06F9/30043, G06F9/3863, G06F12/0888|
|European Classification||G06F9/30A2L, G06F9/38H2, G06F12/08B18, G06F9/38E4, G06F11/14A2C|
|Jun 15, 2005||AS||Assignment|
Owner name: INTERNATINAL BUSINESS MACHINES CORPORATION, NEW YO
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:EISEN, SUSAN ELIZABETH;LE, HUNG QUI;MACK, MICHAEL JAMES;AND OTHERS;REEL/FRAME:016339/0418;SIGNING DATES FROM 20050209 TO 20050210
|Jul 30, 2012||REMI||Maintenance fee reminder mailed|
|Oct 4, 2012||FPAY||Fee payment|
Year of fee payment: 4
|Oct 4, 2012||SULP||Surcharge for late payment|
|Apr 15, 2016||FPAY||Fee payment|
Year of fee payment: 8